AlphaGo doesn’t need any help from humans to master Go, one of the most advanced strategy board games. Earlier iterations of the AI-based system were able to beat the best humans in the world, but it learned to master the game with some assistance from human knowledge.
Now, Google has published a paper explaining how “AlphaGo Zero” was able to learn without assistance.
Along with teaching itself the game from scratch, AlphaGo Zero won 100-0 against an earlier version of AlphaGo, indicating it’s likely the best Go-playing entity in the world. It reached that level of mastery in three days.
